Manic fringe and lunatic fringe modify different sites of the Notch2 extracellular region, resulting in different signaling modulation.

Three mammalian fringe proteins are implicated in controlling Notch activation by Delta/Serrate/Lag2 ligands during tissue boundary formation. It was proved recently that they are glycosyltransferases that initiate elongation of O-linked fucose residues attached to epidermal growth factor-like sequence repeats in the extracellular domain of Notch molecules. Here we demonstrate the existence of functional diversity among the mammalian fringe proteins. Although both manic fringe (mFng) and lunatic fringe (lFng) decreased the binding of Jagged1 to Notch2 and not that of Delta1, the decrease by mFng was greater in degree than that by lFng. We also found that both fringe proteins reduced Jagged1-triggered Notch2 signaling, whereas neither affected Delta1-triggered Notch2 signaling. However, the decrease in Jagged1-triggered Notch2 signaling by mFng was again greater than that by lFng. Furthermore, we observed that each fringe protein acted on a different site of the extracellular region of Notch2. Taking these findings together, we propose that the difference in modulatory function of multiple fringe proteins may result from the distinct amino acid sequence specificity targeted by these glycosyltransferases.

Three mammalian fringe proteins are implicated in controlling Notch activation by Delta/Serrate/Lag2 ligands during tissue boundary formation. It was proved recently that they are glycosyltransferases that initiate elongation of O-linked fucose residues attached to epidermal growth factor-like sequence repeats in the extracellular domain of Notch molecules. Here we demonstrate the existence of functional diversity among the mammalian fringe proteins. Although both manic fringe (mFng) and lunatic fringe (lFng) decreased the binding of Jagged1 to Notch2 and not that of Delta1, the decrease by mFng was greater in degree than that by lFng. We also found that both fringe proteins reduced Jagged1-triggered Notch2 signaling, whereas neither affected Delta1-triggered Notch2 signaling. However, the decrease in Jagged1-triggered Notch2 signaling by mFng was again greater than that by lFng. Furthermore, we observed that each fringe protein acted on a different site of the extracellular region of Notch2. Taking these findings together, we propose that the difference in modulatory function of multiple fringe proteins may result from the distinct amino acid sequence specificity targeted by these glycosyltransferases.
Drosophila and mammalian fringe proteins modulate the formation of compartment border in the developing embryo through affecting Notch activation by the ligands (20 -25). Drosophila fringe inhibits a group of cells from responding to the ligand Serrate and potentiates them to respond to another ligand, Delta (20). In higher vertebrates, three fringe proteins, manic fringe (mFng), lunatic fringe (lFng), and radical fringe (rFng) have been identified (22,26). The mammalian fringe proteins modulate Notch signaling when expressed in Drosophila (26), and lFng null mouse phenotypes are similar to those described for mice deficient in components of the Notch signaling pathway (27,28). A weak sequence similarity shared by Drosophila fringe and a class of bacterial glycosyltransferases predicted that fringe proteins might be a glycosyltransferase (29). In a recent work, it was proven experimentally that Drosophila and mammalian fringe proteins have a fucose-specific ␤1,3-N-acetylglucosaminyltransferase activity that catalyzes the elongation of O-linked fucose on the EGF repeats of Notch (30,31). However, it has not been determined whether these fringe proteins act in a distinct manner on the Notch molecules, and if they do, how the difference is generated.
In this study, we report functional diversity between mFng and lFng, namely a difference in their abilities to modulate Jagged1 binding to Notch2 and in the ligand-triggered Notch2 signaling. Of interest, we also found that the two fringe proteins modify different sites in Notch2, implying that each fringe specifies the multiple putative consensus sequences for O-linked fucose glycosylation in the extracellular region of Notch2.

EXPERIMENTAL PROCEDURES
Preparation of Soluble Fusion Proteins-Soluble DSL proteins (sD1-Fc and sJ1-Fc) were prepared as described previously (32).
Cell-binding Assay-Binding of each soluble DSL protein to the surface of the pro-B cell line BaF3 was performed as describedpreviously (13). Briefly, 3 ϫ 10 5 BaF3 cells were incubated with 6.7 nM soluble DSL-Fc proteins in a cell-binding buffer (PBS containing 2% fetal bovine serum, 100 g/ml CaCl 2 , and 0.05% NaN 3 ) at 37°C after blocking with 5 l of rabbit serum. After a 10-min incubation, the cells were washed three times with the cell-binding buffer and further incubated with a phosphatidylethanolamine-conjugated anti-human IgG antibody. The cells then were analyzed using FACScaliber (Becton Dickinson Immunocytometry Systems).
Coprecipitation Using Soluble DSL Proteins-Coprecipitation experiments using the soluble DSL proteins were performed as described previously (32). Each soluble DSL-Fc (6.7 nM) was allowed to bind to 1 ϫ 10 7 BaF3 cells in a buffer containing 20 mM Hepes (pH 7.5), 150 mM NaCl, and 100 g/ml CaCl 2 . Disuccinyl glutarate (Pierce), a crosslinking reagent, was then added to the soluble DSL-Fc-bound BaF3 at a final concentration of 20 M followed by further incubation for 30 min at room temperature. After the cross-linking reaction, the cells were solubilized in a TNE buffer containing 20 mM Tris-Cl (pH 7.4), 150 mM NaCl, 1.0% Nonidet P-40, 5 g/ml aprotinin, and 1 mM EDTA for 30 min at 4°C. The lysates were precipitated with protein G beads.
Transient Reporter Assay-The transient reporter assay was performed as described previously (32). Chinese hamster ovary ras clone-1 (CHO(r)) cells with or without exogenous full-length Notch2 (N2) were inoculated at 3 ϫ 10 4 in a 24-well plate and transfected with a TP1luciferase reporter plasmid, pGa986 -1, by a liposome-based method (SuperFect, Qiagen). After transfection, the cells were co-cultured for 40 h with 5 ϫ 10 4 of the parental or full-length DSL protein-expressing CHO(r) cell lines. Luciferase activity in the mixture of CHO(r) cells then was measured using a luminometer.

Endogenous and Exogenous mFng Is a Major Determinant
for Binding of DLS Ligands to Notch2-Recently, we showed that soluble Jagged1 (sJ1-Fc) binds less efficiently than soluble Delta1 (sD1-Fc) to Notch2 on a mouse pro-B cell line, BaF3, in a cell-binding assay and a coprecipitation assay (32). To determine the consistency of this relative relationship between Jagged1 and Delta1 in Notch2 binding, we conducted cellbinding assays for other Notch2-expressing cell lines and compared the binding amounts of Jagged1 and Delta1 in various cell lines. The results showed that the lower binding activity of sJ1-Fc observed in BaF3 was maintained in a group of cell lines, such as 32D (mouse myeloid progenitor) and mouse myeloid leukemia cells (designated group A; Fig. 1A). However, sJ1-Fc bound at a level equivalent to sD1-Fc in other cell lines such as CHO(r), fN2-CHO (CHO(r) with exogenously introduced full-length Notch2) and C2C12 (mouse myoblastic) cells (designated group B; Fig. 1A). We then investigated by coprecipitation assay whether the binding of sD1-Fc and sJ1-Fc to the Notch2 receptor expressed on the surface of groups A and B cells was different. As we showed previously, Notch2 fragments derived from a membrane-spanning subunit (N2 TM ) were coprecipitated less efficiently with sJ1-Fc than with sD1-Fc in BaF3 (see figure 3E in Ref. 32). Similar results were obtained with 32D cells (data not shown). In contrast, when fN2-CHO was used instead of BaF3 or 32D, sD1-Fc and sJ1-Fc coprecipitated the N2 TM -derived fragments to a similar extent (Fig. 1B). Results obtained with CHO(r) and C2C12 cells were similar to those with fN2-CHO (data not shown). These results indicate that the binding of Delta1 and Jagged1 to Notch2 varies with the type of cell expressing Notch2 and further suggest that there must be a mechanism that modulates the binding between Notch2 and its ligands.
Speculating that the fringe proteins might contribute to the different pattern in Jagged1 and Delta1 binding to Notch2, we next examined expression of the known mammalian fringe genes by Northern blot analysis. Strong expression of mFng was detected in BaF3, 32D, and mouse myeloid leukemia cells, but it was undetected in C2C12, CHO(r), and fN2-CHO cells. In contrast, lFng and rFng were detected at a similar level in all of these cells (Fig. 2). In a reverse-transcriptase polymerase chain reaction, the expression of mFng was confirmed in BaF3 and 32D but not in C2C12 cells, whereas lFng and rFng were detected in BaF3, 32D, and C2C12 cells (data not shown). These results indicated that the expression of mFng correlated with the differential binding of sD1-Fc and J1-Fc to Notch2. To show the involvement of mFng in the differential binding, we generated fN2-CHO clones expressing exogenous mFng (fN2/ mFng-CHO). At the same time, we also generated fN2-CHO cells expressing exogenous lFng (fN2/lFng-CHO) to see whether there is a functional diversity between mFng and lFng (Fig. 3A). As expected, the cell-binding assays showed that sJ1-Fc bound to fN2/mFng-CHO cells less efficiently than sD1-Fc (Fig. 3B). The fN2/mFng-CHO cells were similar to BaF3 and 32D rather than parental fN2-CHO in that they had a preference for Delta1 (Figs. 1A and Fig. 3B). In contrast, although binding of sJ1-Fc to fN2/lFng-CHO was also less than that of sD1-Fc, the difference was less prominent than that in fN2/mFng-CHO. When the data were transposed, it was revealed that mFng and lFng did not affect or marginally increased the binding of sD1-Fc to Notch receptors on the fN2-CHO cell surface but reduced that of sJ1-Fc (Fig. 3C). Coprecipitation assays then revealed that the amount of sJ1-Fc-coprecipitated N2 TM was significantly less than that of sD1-Fc-cprecipitated N2 TM in fN2/mFng-CHO cells, in contrast to the fact that the amount of N2 TM coprecipitated with sD1-Fc and that with sJ1-Fc were closely similar in parental fN2-CHO cells. In fN2/lFng-CHO cells, the amount of sJ1-Fc-coprecipitated N2 TM was slightly less than that of sD1-Fc-coprecipitated N2 TM (Fig. 3D). These observations indicate that the fringe proteins have the ability to modulate Notch receptors for interaction with their ligands and that each mFng and lFng executes this activity in a different manner.
Modulation of Ligand-activated Notch2 Signaling by Fringe   FIG. 1. Modulation of binding of Notch2 to its ligands. A, two soluble DSL proteins (sD1-Fc and sJ1-Fc) were allowed to bind to various cell lines at a concentration of 6.7 nM. Cell-bound DSL proteins were analyzed by fluorescenceactivated cell sorter. Human IgG (hIgG) was used to determine the background. B, coprecipitation analysis. After incubating the respective DSL-Fc protein with fN2-CHO, protein G beads were added directly to the fN2-CHO lysates to precipitate the DSL-Fc-containing complex. To identify fragments of Notch2, lysates of these cells were precipitated with an anti-Notch2 rabbit polyclonal antibody. These precipitates were analyzed by a Western blot probed with an anti-Notch2 monoclonal antibody. The band with an approximate molecular mass of 120 kDa represents the membrane-spanning subunit (Notch2 TM ). exo N2 TM , Notch2 TM generated by introduced Notch2 cDNA; endo N2 TM , endogenous Notch2 TM expressed in CHO(r); ProteinG, precipitation with protein G alone.
Proteins-We then examined whether the two fringe proteins influenced ligand-triggered Notch2 signaling. In fN2-CHO, the three full-length DSL ligands, Delta1, Jagged1, and Jagged2, activated the transcription of a reporter gene driven by an RBP-J-responsive promoter to a similar extent (see columns designated fN2-CHO in Fig. 4A), as was reported previously (32). Compared with this, Jagged1-activated transcription was decreased in both fN2/mFng-CHO and fN2/lFng-CHO cells (Fig. 4A). The degree of decrease was significantly greater in fN2/mFng-CHO than in fN2/lFng-CHO cells (Fig. 4B), demonstrating a difference in function between mFng and lFng. Del-ta1-and Jagged2-activated transcription was unaffected in fN2/mFng-CHO and slightly enhanced in fN2/lFng-CHO cells (Fig. 4, A and B). When the fringe-generated modulation of signaling through endogenous Notch receptors (Fig. 4D) was subtracted, results indicated that mFng profoundly reduced the Jagged1-triggered Notch2 signaling, and lFng reduced it to a lesser degree. After subtraction, neither fringe significantly modified the Notch2 signaling triggered by Delta1 or Jagged2.
It was reported that the expression of lFng in Delta1-or Jagged1-expressing cells did not affect the ligand-triggered Notch1 activity (33). To determine whether this is also true in the case of Notch2, we separately generated fringe proteinexpressing fD1-CHO and fJ1-CHO cell lines and performed a transient transcriptional activation experiment using these cells as stimulators and fN2-CHO as a host of the reporter gene. The stimulator cells induced constant transactivation of the luciferase gene, irrespective of the exogenous introduction of mFng or lFng (Fig. 5). Furthermore, the addition of mFng or lFng purified from the culture supernatant of mFng-CHO or lFng-CHO did not affect the transcription induced by Delta1 or Jagged1 (data not shown). These observations indicate that the fringe proteins must be coexpressed with Notch2.
Mechanism of Modulation of Notch Signaling by the Fringe Proteins-Next, we investigated whether the modification of the Notch2 molecule by the fringe proteins could be identified, and whether the functional diversity between them could be explained by the difference of such modification. We generated CHO(r) cells expressing either of the three deletion mutants of the extracellular domain of Notch2 with a FLAG tag (15N2-CHO, 22N2-CHO, and 29N2-CHO, for deletion mutants harboring the N terminus through the 15th EGF repeat, through the 22nd EGF repeat, and through the 29th EGF repeat, respectively). We further introduced fringe cDNA into these CHO cells. When the supernatant of each cell line was subjected to the gel and Notch2-FLAG was probed in a Western blot, we found that the migration of the 15N2 derived from lFng-expressing cells was slowed as compared with that derived from the parental 15N2-CHO cells (Fig. 6A). The introduction of mFng or the control vector did not affect the migration of the 15N2 protein. A similar result was observed in 22N2 (Fig. 6B). In contrast, the migration of 29N2 was up-shifted in the presence of either mFng or lFng (Fig. 6C), indicating an increase in the molecular weight of 29N2. These observations indicate that the fringe proteins directly modify Notch2, which is consistent with the recent finding that fringe is a glycosyltransferase that directly modifies Notch (30,31). It was further indicated that lFng does this at a site from the N terminus through the 15th EGF repeat of Notch2, and mFng does so at a site from the 23rd through the 29th EGF repeat of Notch2. DISCUSSION In this study, we characterized the function of mFng and lFng and identified their functional diversities. Both fringe proteins differentially modulated the binding of DSL ligands to the Notch2 receptor as well as Notch2 signaling triggered by DSL ligands. We further obtained biochemical data suggesting that these fringe proteins modify the extracellular region of Notch2 at different sites.
Based on the results of in vivo studies using Drosophila, the notion that fringe inhibits Serrate/Jagged-dependent Notch activation and potentiates Delta-dependent Notch activation has been established (20,21,24). This notion was strengthened recently by in vitro analyses using mammalian versions of these molecules, namely lFng, Notch1, Jagged1, and Delta1 (33). Furthermore, the fringe protein was proved recently to be a glycosyltransferase that modifies the extracellular domain of Notch (30,31). Given the recent finding that Drosophila fringe strengthened the binding of Drosophila Delta to Drosophila Notch (31), it was anticipated that the modification of Notch by fringe would influence binding between the Notch receptor and its ligand. However, it has also been reported that fringe does not affect binding between the Notch receptor and its ligand (30,31,34).
In the present study, the endogenous expression of mFng was clearly correlated to the finding that cell binding of Jagged1 to Notch2 was less than that of Delta1 ( Figs. 1 and 2). It was also shown that the enforced expression of mFng reduced the binding of Jagged1 to Notch2 (Fig. 3D). We therefore suggest that the lower binding of Jagged1 to 32D and BaF3 cells than that of Delta1 resulted from the endogenous expression of mFng, which reduces the Jagged1 binding to Notch2. Furthermore, although mFng did not affect the binding of Delta1 to Notch2 (Fig. 3C), we observed that mFng strengthened the binding of Delta1 to Notch1 (data not shown). It is therefore true that a fringe protein could modify the binding of a DSL ligand to a Notch receptor in selected circumstances if not in all. Whether a certain fringe protein affects the binding of a DSL ligand to a Notch receptor largely depends on the combination of fringe protein, DSL ligand, and Notch receptor involved. The expression of exogenous lFng in CHO cells, which express endogenous lFng, also showed a binding modulatory activity, although in a lesser degree than that in mFng (Fig. 3). Expression level may therefore be a determinant of whether a fringe protein exerts modulatory activity on ligand-Notch binding.
It was reported previously that mFng and lFng inhibited Notch1-mediated signaling triggered by Jagged1 and enhanced that triggered by Delta1, and either Jagged1-or Delta1-triggered Notch2 signaling was enhanced by lFng (33). In contrast, we describe here that mFng and lFng inhibited Notch2-mediated signaling triggered by Jagged1, in which mFng showed stronger activity, and that neither fringe affected Notch2-mediated signaling triggered by Delta1 (Fig. 4). This apparent discrepancy may be because of the difference in the cells that were used to overexpress mFng and lFng. From our results, modification of Notch2 binding to Jagged1 (reduction) and A, the generation of fN2-CHO cell lines expressing FLAG-tagged mFng and lFng proteins, designated fN2/mFng-CHO and fN2/lFng-CHO, respectively. The expression of each fringe protein was verified by Western blot analysis with an anti-FLAG antibody. B and C, the binding of two DSL proteins to fN2-CHO, fN2/mFng-CHO, and fN2/lFng-CHO cells was analyzed by fluorescence-activated cell sorter. The concentration of incubated DSL proteins was 6.7 nM. C, the binding data were re-aligned to show the effect of mFng and lFng on ligand binding to the Notch receptors more clearly. D, coprecipitation analysis. After incubation of 6.7 nM of the respective DSL-Fc protein or control human IgG (hIgG) with fN2-CHO, fN2/mFng-CHO, and fN2/lFng-CHO cells, protein G beads were added directly to the lysates to precipitate an Fc-containing complex. Notch2 TM in each fN2-CHO is shown in the right panel. Two bands represent Notch2 TM generated by introduced Notch2 cDNA (exo N2 TM ) and endogenous Notch2 TM expressed in CHO(r) (endo N2 TM ).
Delta1 (little effect) and modification of Notch2-mediated signaling triggered by Jagged1 (inhibition) and Delta1 (no effect) were correlated with each other (Figs. 3 and 4). It is not clear, however, whether this correlation is of biological significance or represents simple coincidence.
To date, the question of functional diversity among the fringe proteins has been poorly addressed. Although a difference has been shown among Drosophila fringe proteins, mFng and lFng, in their abilities to transfer GlcNAc to O-linked fucose on an EGF repeat of serum protein factor VII (30), the significance of the difference has remained unclear, because serum protein factor VII is not a physiological target for fringe. As described above, we revealed a functional difference between mFng and lFng in their ability to modulate Jagged1 binding to Notch2 and Jagged1-triggered Notch2 signaling (Figs. 3 and 4). In addition, we found that lFng modifies at a site from the N terminus through the 15th EGF repeat of Notch2, and mFng acts at a site from the 23rd through the 29th EGF repeat (Fig.  6), suggesting that each fringe has its specific site for the modification of Notch2. This could explain why the two fringe proteins cause the different modulation of Notch2 signaling. We note that 5 of 12 O-linked fucose consensus sites (2nd, 3rd, 5th, 6th, and 8th EGF repeats) are indeed present within the N-terminal through the 15th EGF repeat, and three consensus sites (24th, 25th, and 26th repeats) exist within the 23rd through the 29th EGF repeat. The latter site, representing an mFng target region, coincides with the Abruptex domain of Drosophila Notch comprising EGF-like repeats 24 -29, which was identified recently as a domain that may mediate interactions between Notch, its ligands, and fringe (35,36). Given the fact that Drosophila fringe and Notch interact with each other (36), we suppose that the specificity may be determined by the difference in the site(s) of Notch2 with which the fringe proteins interact.
In the present study, we reveal the existence of functional diversity among the fringe proteins. Given that it is known that the strength of Notch signaling is critical for the exact cell fate decision based on the fact that abnormal phenotypes are exhibited by haploinsufficiency of DSL ligand (37,38), we propose that multiple fringe proteins with diverse signal-modulation activities exist in higher vertebrates to rigidly control the strength of Notch signaling. We also show that Jagged2-triggered signaling is much less affected by either fringe protein than Jagged1-triggered signaling, despite significant structural similarities between the two ligands. We believe that these findings will facilitate our understanding of the complexities of Notch signaling in higher vertebrates, in which multiple Notch receptors, DSL ligands, and fringe proteins exist.
FIG. 5. Ligand-induced Notch signaling is not affected by exogenous fringe, which is expressed in stimulator CHO cells. We addressed the question of whether fringe proteins could exhibit modulating activity for ligand-induced Notch signaling when they are introduced into the stimulator CHO cells with exogenous Notch ligands, as was observed on their introduction into the fN2-CHO cells used as a target (Fig. 4A). To clarify this, we generated respective fringe proteinexpressing fD1-CHO and fJ1-CHO cell lines (fD1-mFng-CHO, fD1-lFng-CHO, fJ1-mFng-CHO, and fJ1-lFng-CHO) and performed a transient transcriptional activation assay using the generated cells for stimulation and pGa986 -1-transfected fN2-CHO as the target.
FIG. 6. Direct modification of the Notch2 receptor by fringe proteins. The supernatants of the three kinds of sN2-expressing CHO cells, with or without each fringe cDNA, were subjected to Western blot analysis probed with an anti-FLAG antibody. A, sN2 (EGF1-15) comprising the N terminus through the 15th EGF; B, sN2 (EGF1- 22) comprising the N terminus through the 22nd EGF; C, sN2 (EGF1-29) comprising the N terminus through the 29th EGF. The asterisk indicates a positive band shift.